Insights into binary evolution from gravitational waves

Transcription

1 Insights into binary evolution from gravitational waves Simon For the COMPAS team Alejandro Vigna-Gomez, Jim Barrett, Coen Nijssell, Christopher Berry, Ilya Mandel and more

2 GW LVC 2016 arxiv:

3 GW LVC 2016 arxiv:

4 Masses Stevenson+ 2015, Stevenson (in press), Mandel

5 Measured masses Combination of component masses: Chirp mass for low mass events Total mass for high mass events SAMSI Astro telecon LVC 2016 arxiv: /11/2016 5

6 BBH (total) mass distribution Assumes our Fiducial model values of hyperparameters, integrated across cosmic history Data from Stevenson in press Evolution of BBH merger rate over cosmic history studied in Neijssel in prep SAMSI Astro telecon 10/11/2016 6

7 Progenitors of O1 events (maybe) Stevenson et al 2017 in press 7

8 Typical evolutionary channel for GW Time M 1 ST 1 ST 2 M 2 [Myr] [M ] [M ] [R ] MS MS HG MS HeMS MS BH MS a BH CHeB BH CHeB BH HeMS BH BH Stevenson et al 2017 in press 8

9 Uncertainties in binary evolution Initial conditions (Recent progress e.g. Sana+ 2012, probably correlated as in Moe+ 2016) Stellar evolution (in particular for massive stars e.g. expand or not? Overshooting -> core masses -> BH masses) Supernovae NS/BH birth kicks same or different? BH kicks large or small? NS/BH mass spectrum? Spin tilts? Stellar winds absolute mass loss rates, extrapolation in metallicity Mass transfer e.g. accretion efficiency, mass loss mode, response to mass loss (stable v unstable) Common envelope? Structure of massive stars and response to mass loss? Efficiency of common envelope ejection? 9

10 Can be explored with pop synth Figure from Stevenson

11 Model comparison to pop synth models Currently only include small set of isolated binary evolution models (Dominik et al 2012) would like to include other channels Figure from Stevenson

12 After O1 Unpublished, after Stevenson

13 Chirp mass changes with hyper parameters of pop synth models Barrett et al 2017 in prep 13

14 Interpolate chirp mass distribution Previous work by O Shaughnessy in interpolating pop synth rate in high dimensions We bin chirp mass distribution, calculate principle component analysis and then interpolate coefficients using Gaussian Process Similar method to Taylor for PTAs ( ) Barrett et al 2017a, Barrett et al 2017b in prep 14

15 e.g. e.g. kick velocity Concordance cosmology binary evolution Use interpolated pop synth models to solve inverse problem; Which combination of hyperparameters in our model best explains the Gravitational waves? gravitational wave observations? Figure adapted from Kowalski et al e.g. common envelope efficiency 15

16 Model independent methods Mandel Stevenson

17 Mass function with number of observations Mandel Stevenson

18 Spins Stevenson in prep (see also Salvo s talk earlier) 18

19 Models for black hole spin-orbit misalignment angles We use a simplified population synthesis code (COMPAS) to model the binary black hole population We vary our assumptions about spin-orbit misalignments For all of our models we assume: The magnitude of both black hole spins is 0.7 Black holes receive linear kicks in a similar way to neutron stars during a supernova The mass distribution is identical for all channels All binaries form with spins aligned to orbital angular momentum 19

20 Isolated binary evolution 1) ˆL Many (but not all) binaries formed spin-aligned Many astrophsyical processes (tides, mass transfer etc) act to realign spins with the orbital angular momentum We assume that both black holes are aligned when they merge Possible if no kicks in BH formation and stars form aligned Also have 2 additional models 3) and 4) that vary assumption of both spins being exactly aligned 20

21 Dynamical formation 2) ˆL We assume that the binary black hole is formed dynamically Both spins are misaligned isotropically and uncorrelated (e.g. Rodriguez+ 2016) and remains isotropic into LIGO band Possible for both stellar dynamics e.g. globular clusters and dynamically formed primordial black hole binaries 21

22 Measuring misaligned spins with GWs Aligned spins Misaligned spins M 1 = M M 2 = 7.52 M a 1 = a 2 = 0.7 Calculate for Advanced LIGO at design sensitivity 22

23 Model 3 Isolated binary evolution 3) Prior to 2 nd Supernova ˆL After 2 nd Supernova ˆL We assume both aligned prior to the second supernova via CE Since misalignment set by second supernova kick, both BH spins are typically only modestly and equally misaligned, causing them to freely precess (Kalogera 2000, Schnittman 2004, Apostolatos 1994) 23

24 Model 4 Isolated binary evolution 4) Prior to 2 nd Supernova After 2 nd Supernova ˆL ˆL We assume the secondary is aligned prior to the second supernova via tides, with the primary misaligned via a supernova kick After second supernova, primary can be misaligned by a large angle, secondary by a more modest one This leaves the primary misaligned and secondary aligned (as in Gerosa et al 2013). 24

25 Mixture model - distributions of black hole spin misalignments 1) 2) 3) 4) 25

26 Hierarchical analysis Our models overlap significantly in parameter space. The spin-orbit misalignment angles are poorly measured for individual events A la Hogg 2010, Mandel 2010, we sample from the posterior given by the likelihood: Introduced by Chris earlier 26

27 How constraint on fraction of dynamically formed BBHs evolves with the number of observations Including realistic measurement uncertainties Drawing increasing number of observations from a multinomial distribution True fractions shown in blue 27

28 How constraint on fraction of dynamically formed BBHs evolves with the number of observations Dark shaded region is 1 sigma, light is 2 sigma Roughly 1/sqrt(nObservations) in the tail 28

29 Do misaligned spins really correspond to different formation channels? Possibility of spin tilts in supernovae (a la double pulsar Farr et al ) cause binaries to lose memory of formation. From a modelling point of view: Can we relate pre-sn stellar spin to post-sn BH spin? How well do we understand realignment in e.g. common envelope? 29

30 Conclusions Isolated binary evolution is highly uncertain corresponding to uncertainty in supernovae and mass transfer (inc common envelope evolution) Gravitational waves provide a way to probe binary evolution and we now have observations! Can determine fractions of systems coming from isolated binary evolution v other channels Can use observations of GW masses, spins and rates to place constraints on astrophysical hyperparameters which go into our model, corresponding to uncertain astrophysics 30